Research methods and techniques — OCR A-Level Psychology
Test yourself on Research methods and techniques with OCR A-Level practice questions.
A week of Premium from your first lesson · Then the free plan continues · No card, no charge
Research methods and techniques explained
This is assessed as a skill rather than as a vocabulary test, so each technique has to be held three ways at once: what it is, when a researcher would choose it, and what choosing it costs.
Read the full explanation
The trade off that organises the whole section is control against realism. Milgram (1963) ran his obedience procedure in a Yale laboratory, which let him standardise the prods, the shock generator and the learner's scripted protests, but bought that precision with a task nobody meets in ordinary life. Piliavin et al. (1969) staged a collapse on a moving New York subway train and recorded helping that nobody knew was being studied, while losing control over who happened to be in the carriage. Marks go to candidates who pick the strength or weakness that actually bites on the scenario in front of them, so revise every method with one real study attached to it.
Experiment: laboratory experiment; field experiment; quasi experiment.
All three count as experiments because a dependent variable is measured across levels of an independent variable; what separates them is the setting and who decided which condition a person was in. Loftus and Palmer (1974) changed a single verb in a question about a filmed collision, and mean speed estimates rose from about 31.8 mph for contacted to about 40.8 mph for smashed, a difference that reads as causal only because everything else was held constant. Piliavin et al. (1969) staged a collapse on a subway train, so the helping measured was real behaviour, but the make up of each carriage varied from trial to trial. Casey et al. (2011) compared adults who had delayed gratification as children with those who had not; that grouping existed long before the study began, so nothing in the design shows the delay caused the later difference.
Observation: structured; unstructured; naturalistic; controlled; participant; non-participant; overt; covert.
Observational work records what people do instead of asking them, and the eight labels here are really four independent decisions: how much structure the recording has, how much the situation is arranged by the researcher, whether the observer joins in, and whether the people know. Bandura, Ross and Ross (1961) sits at the structured and arranged end, since nursery children were watched through a one way screen and scored in short time intervals against fixed categories such as imitative physical aggression and mallet aggression, which is what made agreement between two observers measurable. Piliavin et al. (1969) sits at the hidden end, with observers seated in the same carriage recording who helped and how long they took, and nobody told. The gain each time is behaviour unshaped by being asked about; the cost is either artificiality or consent.
Self-report: questionnaire; Interviews: structured, semi-structured, unstructured.
Asking people reaches what watching them cannot, namely reasons, beliefs and private experience, and the whole evaluation turns on whether the respondent can and will tell you. Closed items and rating scales give numbers that compare easily across a large sample, at the cost of forcing an answer nobody actually holds; open items and a loose interview give depth that resists summary. Kohlberg (1968) put moral dilemmas to boys and scored the reasoning behind each answer rather than the answer itself, which needed an interviewer free to follow up. Hancock et al. (2011) asked convicted murderers to describe their offence in their own words, then ran the transcripts through text analysis software, finding that those scoring high on psychopathy used more cause and effect connectives and more talk of basic needs such as food and money. Social desirability threatens both.
Correlation: obtaining data for correlational analysis; positive correlation; negative correlation; no correlation.
Here the researcher measures two things that already vary and asks how they vary together, manipulating nothing, which is exactly why no causal claim can follow. The data must be pairs taken from the same participant, each pair plotted as one point on a scattergram, and the coefficient that summarises the cloud runs from minus one through zero to plus one. Maguire et al. (2000) is the example worth carrying: the longer London taxi drivers had spent doing the job, the greater the grey matter volume in the posterior hippocampus and the smaller it was in the anterior region, so one study holds a relationship in each direction. Neither shows that driving changed the brain, since people with a larger posterior hippocampus may simply last longer in the work, and setting out that alternative is the evaluation examiners reward.
Your focus
- Apply appropriate research methods and techniques to novel scenarios by identifying manipulated variables, settings, or measures.
- Compare the strengths and weaknesses of different research methods, focusing on the trade-off between control over extraneous variables and ecological validity.
- Evaluate research methods by using evidence from named core studies to assess issues such as demand characteristics and ethical considerations.
Show all 15 objectives
- Apply knowledge of experimental designs to identify independent and dependent variables within a scenario, explaining how the dependent variable is operationalised.
- Analyse the differences between laboratory, field, and quasi-experiments in terms of setting, allocation of participants, and the ability to establish causal relationships.
- Evaluate the use of different experimental methods by weighing the reduction of demand characteristics against the loss of control and ethical issues surrounding informed consent.
- Explain the different types of observational methods by categorising them according to structure, setting, observer participation, and participant awareness.
- Apply principles of structured observation to design observable, non-overlapping behavioural categories and select appropriate time or event sampling rules.
- Evaluate observational designs by assessing inter-rater reliability and weighing the internal validity gained from covert observation against ethical costs such as lack of informed consent.
- Compare structured, semi-structured, and unstructured interviews in terms of researcher control and the type of qualitative or quantitative data they produce.
- Apply principles of self-report design to construct clear, non-leading questions and appropriate response formats, such as rating scales.
- Evaluate the validity of self-report measures by analysing the impact of social desirability bias on participant responses in specific research contexts.
- Explain how data for correlational analysis is obtained and plotted on a scattergram with appropriately labelled axes and one point per participant.
- Analyse the strength and direction of relationships between co-variables to identify positive, negative, and zero correlations.
- Evaluate correlational research by explaining why causal conclusions cannot be drawn, referencing the third variable problem and bidirectional ambiguity.
Research methods and techniques exam tips
Marking Points
- Naming the method scores little on its own; credit comes from identifying it out of the stem, by pointing to a variable the researcher manipulated, a setting the participants already belonged to, or a measure taken without any manipulation at all.
- A strength or weakness is only credited when it is tied to the study described, so write that this standardised procedure could be repeated by another researcher rather than writing that laboratory experiments are replicable.
- Comparisons earn marks when the dimension being compared is named: control over extraneous variables, ecological validity, demand characteristics, or the ethics of consent and deception.
- Using a named study such as Milgram (1963) or Piliavin et al. (1969) as evidence about a method shows knowledge and evaluation in the same sentence, which is what the higher mark bands ask for.
- Identify the independent and dependent variables in the words of the scenario, and say how the dependent variable was operationalised, because memory is not a measure whereas number of items correctly recalled is one.
- A study is quasi when the independent variable is a pre-existing characteristic such as age, sex, diagnosis or expertise, so the researcher allocates nobody and a causal claim cannot be made.
- Field experiments gain credit for reduced demand characteristics, since participants do not know they are taking part, and lose it for weaker control and for consent that cannot be sought beforehand.
- State what would have to be true for causation, namely manipulation of one variable plus control of the others, then judge the described study against that standard.
- Say which of the paired decisions apply and treat each separately, because a single study can be covert, unstructured and participant all at once.
- Structured work earns credit for behavioural categories that are observable and do not overlap, and for the sampling rule used, whether time sampling at fixed intervals or event sampling of every occurrence.
- Inter-rater reliability is the evaluation examiners look for: two observers coding the same behaviour independently, then their records compared or correlated.
- Covert work buys internal validity by removing demand characteristics and loses ground on informed consent, privacy and the right to withdraw, and both halves need saying.
- Separate the three interview types by who controls the questions: a fixed schedule read in the same order, a schedule with freedom to probe, or a topic and no schedule at all.
- Credit comes from linking format to data, so closed items give quantitative data that can be compared across a large sample while open items give qualitative detail no fixed option would have caught.
- Name social desirability bias and say what it would do to this measure in particular, for example inflating reported medication use or reported willingness to help.
- A well written item carries one idea, no double negative, no leading wording, and a stated response format such as a five point scale.
- State the direction using the two variables themselves, as in the more hours slept the fewer errors made, rather than announcing that a negative correlation was found.
- A scattergram needs both co-variables labelled on the axes, one point per participant, and no line drawn through the points.
- Strength and direction are separate judgements: a coefficient close to zero means almost no relationship, while a coefficient of 0.8 is strong whether its sign is positive or negative.
- The third variable problem and the possibility that the causal arrow runs the other way are the two criticisms that reliably earn evaluation marks.
Examiner Tips
- 💡Questions on the research methods paper usually open with an unseen scenario and ask you to identify the method and then explain one strength and one weakness of using it here, so read the stem twice and underline what was manipulated and what was measured.
- 💡Evaluation marks are awarded per developed point rather than per point named, so two criticisms taken through to their consequence beat five listed ones.
- 💡The command word suggest tells you the marks are for applying methods knowledge to an unfamiliar situation, so write about that situation rather than about methods in general.
- 💡A recurring question gives an unfamiliar study and asks why it is a quasi experiment; the answer is always about who or what decided which condition each participant was in.
- 💡Asked to suggest how a study could be improved, name a change to a control, a measure or the recruitment, then say what problem the change removes.
- 💡If told to compare field and laboratory experiments, structure by dimension rather than by method: control first, then realism, then ethics.
- 💡You are often asked to write two or three behavioural categories for an unfamiliar scenario at one mark each, and the mark is lost for anything a second observer would have to interpret.
- 💡Asked to explain an advantage of covert observation in a given study, name the demand characteristic that would have appeared had the participants known.
- 💡Questions that ask you to suggest a way of gathering the data usually want a coding schedule or a sampling method, so state which and give the interval or the event.
- 💡You may be asked to write one closed and one open question on a given topic and then explain an advantage of each, so practise producing both quickly with the response format written in.
- 💡A question about handling interview data usually wants thematic analysis described as coding transcripts into recurring themes, plus one problem of researcher subjectivity.
- 💡Where the command word is justify, the marks are for the reason your format fits the aim, not for restating which format you picked.
- 💡Correlations turn up in the data handling questions, so be ready to sketch a labelled scattergram, describe what it shows in one sentence, and name the test you would run.
- 💡Spearman's rho is the test to name for ranked or ordinal co-variables, and you then calculate the value and compare it with the critical value for the number of pairs at the five per cent level, one tailed only if direction was predicted.
- 💡A common four mark item gives you a coefficient and asks what it tells the researcher, so answer with direction, then strength, then the warning about causation.
Common Mistakes
- Treating ecological validity as a property of the setting alone, so a study done in the street is called high in ecological validity even when the task is contrived.
- Offering evaluation that would fit any study at all, such as a bare claim that the findings cannot be generalised, with nothing about what in this task fails to transfer.
- Confusing reliability with validity, then offering the fact that a procedure can be repeated as evidence that its findings are true.
- Calling any study done outside a laboratory a field experiment even when nothing was manipulated; with no independent variable it is an observation or a correlational study.
- Claiming a quasi experiment demonstrates cause and effect because it has two conditions and a significant difference between them.
- Treating laboratory as a synonym for artificial task; the setting and the task are two separate judgements, and a laboratory study can use a lifelike task.
- Writing categories that call for interpretation, such as aggressive or anxious, which two observers will score differently; a usable category names what the body does.
- Confusing participant observation with a participant in a study; the term describes the observer's role, not the person being watched.
- Assuming naturalistic means unstructured, when a researcher can use a rigid coding schedule in a school playground.
- Treating high agreement between observers as proof that the observation was valid, when it only shows it was consistent.
- Calling a questionnaire unreliable when the point being made is about honesty, which is a validity problem rather than a consistency one.
- Describing a structured interview as identical to a questionnaire; the difference is the live interviewer, who introduces interviewer effects but can clarify a misread question.
- Writing a leading item into an answer that was supposed to demonstrate good question design, then losing the mark on the wording alone.
- Offering the bare claim that people may lie, with no account of why they would lie in this particular study.
- Writing that the study proves one variable causes the other, or sliding into causal verbs such as affects, leads to and produces while describing a correlation.
- Reading a negative correlation as no relationship or as a weak one, when the sign gives direction and only the size gives strength.
- Drawing a bar chart for correlational data, or joining up the points on a scattergram as though it were a line graph.
- Taking a significant result in a very large sample as proof of a large effect.